You are on page 1of 7

Civil and Environmental Research, Vol.

5 2013
Special Issue for International Congress on Materials & Structural Stability, Rabat, Morocco, 27-30 November 2013

Seismic Fragility and uncertainty Analysis of Concrete Gravity Dams


under Near-Fault Ground Motions
Hebbouche Abdelhamid1, Bensaibi Mahmoud2 and Mroueh Hussein3
1
MVRE, cole Nationale Suprieure dHydraulique, Blida-Algrie
2
Department of civil engineering, Saad dahlab university, Blida, Algeria
3
Universit Lille 1 Sciences et Technologies, France

Abstract. Throughout the world there are numerous concrete gravity dam has been made in areas of very high
seismicity with least attention to seismic actions. Dam safety during and after an earthquake, is the objective of
the present study. The failure of a dam in a seismic excitation has dramatic consequences in terms of loss of
human lives and financial losses. In the present work, an analytical fragility analysis was performed in order to
characterize the seismic vulnerability of concrete gravity dams by using a numerical simulation procedure to
model sources of uncertainty that could impact dam performance, with combination with non-linear dynamic
response analysis. The seismic fragility of concrete gravity dams under near-fault ground motions was
performed and compared to assess their performance against seismic hazards. An uncertainty analysis is also
carried out to evaluate the effectiveness of the Latin Hypercube Sampling method using different combinations
of performance thresholds through fragility analysis. A case study was considered, it is about the dam of Oued
el Fodda on the Oued Chelif River, West Algeria. This dam was designed in the early 1930s.

exceed (152 cm/s) or, the distance of the epicenter of the


earthquake should be within approximately 15 km of a
structure of interest [16]. A sensitivity analysis is also
carried out to evaluate the effectiveness of the Latin
Hypercube Sampling method.
The aim of this study is to develop the analytical seismic
fragility curves of concrete dams subjected to near-fault
ground motion excitations. The methodology is applied
on the Oued el Fodda dam located west Algeria. It was
built in the early of 1930s.

1 Introduction
Many damaging earthquakes have been recorded in
Northern Algeria (Chelif 1980, Mw=7.3; Ain
Temouchent 1999, M=5.7; Beni ourtilane 2000, M=5.6
and the one in Boumerdes 2003, Mw=6.8), indicating the
importance of the seismic hazard assessment for this
region [1-3]. The seismic fragility that describe the
probability of a structure being damaged beyond a
specific damage state for various levels of ground
motions. In the literature, several studies for generating
seismic fragility curves have been developed [4-13].
Fragility curves are plots of system fragilities versus a
scalar measure of seismic intensity. Traditionally, peak
ground acceleration (PGA) has been used as an intensity
measure. Recent studies show that pseudo-spectral
acceleration provides a superior measure of seismic
intensity than PGA [14].
In recent years, many efforts have been devoted to the
characterization of effect of near-fault ground motions on
the structures. Near-fault ground motions have caused
much damage in the vicinity of seismic sources during
recent earthquakes (Northridge 1994, Kobe 1995 and
Taiwan 1999). There is evidence indicating that ground
shaking near a fault rupture may be characterized by a
short-duration impulsive motion that exposes structures
to high input energy at the beginning of the record
[15]. There are two factors for classifying ground motions
as near-fault. The spike in velocity should generally

2 Structural modeling of dam behavior


The study is based on a concrete gravity dam with a
vertical upstream face, which maintains a reservoir of
water that extends to infinity in the upstream direction
and is based on a semi-infinite foundation. The geometry
of the dam-reservoir-foundation system is shown in
figure.1.

123

Civil and Environmental Research, Vol.5 2013


Special Issue for International Congress on Materials & Structural Stability, Rabat, Morocco, 27-30 November 2013

tensile strength is assumed to be 15% of compressive


strength (3.75 MPa). The water has the unit weight, 1000
kg/m3, pressure wave velocity 1440 m/s. In the analysis,
the damping ratio is assumed to be 5% of the
fundamental frequency of system.
Table 1 lists the near-fault earthquake records selected to
create an ensemble for the seismic fragility of the Oued el
Fodda Dam. All occurred between 1987 and 1999, and
have epicentral distances of 4.77 to 10.36 km with
magnitudes ranging from 6 to 7.1.
Fig. 1. Geometry of dam-reservoir-foundation system

Figure. 3 show the spectral accelerations used for scaled


the both near-fault earthquakes used in the time history
analyses.

2.1 Description of model


The model of the dam-foundation system is illustrated in
figure.2, used 4-node, bilinear finite elements. The
concrete-rock interface is assumed to be horizontal and to
obey the Coulomb friction law. The foundation material
was assumed to be a Mohr-Coulomb material, with its
non-linear behavior assumed to be perfectly plastic. The
concrete in the dam was modeled as an impervious
material. The bottom horizontal boundary of the FE
model is the application point of the de-convolved
seismic ground motion. Different boundary conditions
must be imposed on the nodes on the vertical boundaries
of the FE model. Those nodes, representing the outlying
nodes where the effect of damfoundation interaction is
presumed to have attenuated, are constrained to move
together in the horizontal direction. While the spatial
variation of the earthquake ground motion across the base
of the model is neglected, those nodes in the vicinity of
the base of the dam clearly are affected by the dam
foundation interaction. The above provisions provide an
adequate model for the damfoundation interaction.

Table 1. Properties of selected Near-fault earthquakes records


[17]
N

Earthquake

Year

Magnitude

E.D
[km]

PGA
[m/s2]

S1

Cap
Mendocino

1992

7.1

10.36

1.497

S2

Cap
Mendocino

1992

7.1

10.36

1.039

Kocaeli
Izmit
Turkey

1999

7.4

5.31

0.152

S4

Kocaeli
Izmit
Turkey

1999

7.4

5.31

0.22

S5

Whittier
Narrows

1987

4.77

0.304

S6

Whittier
Narrows

1987

4.77

0.199

S3

Fig. 3. Spectral acceleration of Near-fault earthquakes records


Fig.2. Finite element model of dam-foundation system

3 Structural fragility model


2.2 Material properties

The fragility modeling process allows the combined


effect of the uncertain variables to be propagated through
the model by numerical means (e.g. simulation). The
fragility is modelled commonly by a lognormal
cumulative distribution function (CDF) [4,5].

In this study, the values of material properties used for


dam model are: Unit weight of the concrete is 2500
kg/m3, 0.2 Poissons ratio, and the modulus of elasticity
is taken as 31000 MPa. Compressive strength of the
concrete has been assumed as 25 MPa. The concrete

124

Civil and Environmental Research, Vol.5 2013


Special Issue for International Congress on Materials & Structural Stability, Rabat, Morocco, 27-30 November 2013

F ( )=

Table 2. Uncertain parameters

(1)
Variables

Where [ ] = the standard normal probability integral,


mR=median capacity (expressed in units that are
dimensionally consistent with the demand parameter, y,
e.g. spectral acceleration) and C, the combined
uncertainty is:
=

(2)

Where, R is the logarithmic standard deviation


describing the inherent (aleatory) uncertainty and U is
the logarithmic standard deviation describing the
epistemic uncertainty.

Var 1

Angle of friction

Var 2

Cohesion

Var 3

Dilation angle of
foundation

Var 4

Young modulus of
concrete

Var 5

Young modulus
of soil

Var 6

Compressive strength
of concrete

U (34; 45) degrees


U (0.145; 0.435) MPa
U (27; 33) degrees
U (31.2; 36) 103 MPa
U (40; 80) 103 MPa
N (35;4.8) MPa

These values are based on an review of data summarized


for various types of intact rock [20-23].

The steps for constructing the analytical fragility curves


are as follows:
-

Probability
distribution

Random variables

3.1 Treatment
estimates

Select the earthquake ground motion records;


Scaling ground motion records to the same spectral
acceleration at the fundamental frequency of
structure;
Make an analytical model of the structure;
Modeling uncertainty with Latin Hypercube
Sampling method;
Select uncertainty parameters;
Perform the non linear dynamic response analysis
using selected records and uncertainty parameters;
Construct the fragility curves using the obtained
response and the ground motion indices for each
limit state.

of

uncertainty

in

fragility

Seismic fragilities that incorporate sources of uncertainty


considered above can be derived efficiently using Latinhypercube sampling (LHS) [24] coupled to the finite
element structural models. LHS is a stratified sampling
procedure in which the PDF of each input variable, Xi,
i=1,,k, is divided into N disjoint intervals of equal
probability. Latin Hypercube Sampling (LHS) provides a
stratified sampling scheme rather than the purely random
sampling, providing a more efficient means for covering
the probability space than Monte Carlo simulation
[25,26]. The sampling plan is given by
S = (P R)

Uncertainty modeling
A number of sources of uncertainty are present in the
modeling of fragility of the dam-foundation system and
have been described statistically. Concrete compressive
strength is assumed to be described by normal probability
distributions [19]. For our study, the usefull statistical
data are limited. Therefore, a uniform distribution was
chosen to model the remaining variables. These
parameters are taken as follows table 2:

(3)

where P is an N K matrix, in which each of the K


columns is a random permutation of 1, 2, . . . ,N; R is an
N K matrix of independent random numbers from the
uniform distribution U(0, 1); and N and K are the
numbers of hypercubes and uncertain parameters,
respectively [25,26]. Each element of S, sij, is then
mapped according to

= !"$# (% )

(4)

Where (F-1) is the inverse of cumulative distribution


function (CDF) for parameter j. Each row of x contains
different sets of sampled parameters, from which
statistical samples were obtained.
4 Fragility analysis
The seismic fragilities for the Oued el Fodda Dam are
developed from non linear dynamic analyses and to get a
wider range of the variation of input ground motion,
strong motion records were selected. This last are
conducted with a set of earthquakes include six near-fault
ground motions that are scaled to different spectral
acceleration levels which varying between 0.2g and 2g

125

Civil and Environmental Research, Vol.5 2013


Special Issue for International Congress on Materials & Structural Stability, Rabat, Morocco, 27-30 November 2013

with a step of 0.2g. However, every ground motion set


has different spectral acceleration at the fundamental
frequency of Dam. Six finite elements analyses were
performed for each randomized group and the results
were adequately treated. All results presented herein will
be discussed for a scenario of a strong ground motion
with a spectral acceleration of 2.0g. For each limit state,
three performance measures and corresponds fragilities
are presented.
For LS1 the fragility curves are shown in figure 4. The
probability of exceedance indicating that for tensile
stresses in the neck of the dam greater than 1.0 MPa , 1.5
MPa and 2.0 MPa are 100%, 76.01% and 32.57%,
respectively.

The fragilities for the displacement of the top of the


dam (LS3) with respect to the heel are depicted in
figure 6. A seismic excitation with spectral
acceleration of 2.0g would cause relative
deformations of 5 mm; 20 mm and 40 mm with
probabilities of 100%; 91.32% and 82.45%
respectively.

Figure 6. Seismic Fragility Curves for LS3


Displacement at the top of dam

Figure 4. Seismic Fragility Curves for LS1


Tensile stress at the neck of dam

Figure 7. Seismic Fragility Curves for LS4


Compressive stresses at the heel of dam

These deformations are very small compared to the


height of the dam and only minimal damage should be
expected to gates and other appurtenant structures and
operating equipment due to these deformations.
Moreover, the fact that the overall deformations are on
the order of 0.0004% of the height of the dam suggests
that a rigid body model of the monolith might be an
appropriate simplification to the problem, provided that
one is not interested in the likelihood of tensile cracking
at the neck of the dam.

Figure 5. Seismic Fragility Curves for LS2


Sliding at Dam-Foundation interface

The fragility curves for sliding (LS2) at the


damfoundation interface in figure 5. Indicate that
probabilities of sliding 5mm and 10 mm are very
high, while the probability of sliding 20 mm is about
43.3 %. Thus, some damage to the drainage system,
particularly at the damfoundation interface, might
be expected at this intensity of seismic excitation.

Limit stat 4 is related to material failure and it was


achieved if stresses at the heel of the dam exceeds the
compressive strength of the concrete (25 MPa). It was
found that the fragility illustrated in figure 7. Show that a
compressive stresses of 1 MPa, 3 MPa and 6 MPa had a
probability of failure of 100 %, 89.98 % and 39.28 %
respectively.

Relatively large sliding 20 mm or more could cause


differential movements between adjacent monoliths
in the dam and initiate monolith instability leading
to eventual loss of pool control.

5 Quantification of uncertainties
126

Civil and Environmental Research, Vol.5 2013


Special Issue for International Congress on Materials & Structural Stability, Rabat, Morocco, 27-30 November 2013

In order to identify the principal sources of uncertainties


in the sensitivity analysis, regression analyses were
performed for two cases in this study. In general, in these
regression analyses, the dependent variable is risk
expressed in terms of various uncertain parameters. These
analyses are particularly useful in investigating how
uncertainties in source term variables affect the responses
of the structural system. Also determined were partial
correlation coefficients that represent the importance of
uncertain variables as a function of the magnitude of the
environmental risk. To confirm the influence of each
variables parameters as well as to remove the covariates
on the correlation between a given input variable and the
response variable, a sensitivity analysis based on the
partial correlation has been performed. The partial
correlation coefficient (PCC) between two random
variables Xi and Y given a set of covariates
is defined as
X / i = {X 1 ,..., X i 1 , X i +1 ,..., X p

Figure 8. Effects on the displacement at the top of dam

follows Equations 5 to 9 [27,28]:


&'' = () *+,-.,/- , +1.,/- 2

+,-.,/- = 3 34 5
34 5 = 67 +

+1.,/- = : :4
:4 = ;7 +

6 .3

7
8

; .3

Figure 9. Effects on the Compressive stresses at the heel of dam

9
It can be shown that for both sensitivity cases studied the
cohesion (Var2); the young modulus of concrete (Var4);
the young modulus of soil (Var5) and the compressive
strength of concrete (Var6) are the most influent when
the dam is subjected to near fault earthquakes. As the
displacement response and compressive stress are directly
dependent on Var2; Var4; Var5 and Var6, the sensitivity
should increase as shown in figure 8. And figure 9.

With:

, are the reliability index; X , Y are the random


i
variables

is

and
the

X \i
partial

is

the

correlation

covariates;
coefficients;

e X i . X / i is the residual of prediction of Xi by X\i


eY . X / i is the residual of prediction of Y by X\i ; X i & Y
are the regression variable.

On the other hand, the Var1 and Var2 have a minimal


sensitivity defined by the displacement at the top and
stress at the heel of dam.

The effect of the different input parameters on the


displacement at the top of dam and the compressive
stresses at the heel of dam under near faults earthquakes
using the PCC are given on figures 8 and figure 9.

Conclusions
The evaluation of seismic fragility curves of dams
involving damreservoirfoundation interaction is
studied in this paper. The seismic fragility curves were
studied by means of numerical simulation procedure
Latin Hypercube Sampling (LHS) in combination with
nonlinear dynamic analysis. Concrete gravity dams
subjected to near-fault earthquake scaled to different
spectral acceleration were calculated; results for spectral
acceleration of 2.0g have been considered and discussed.
A series of potential sources of uncertainty associated
with a seismic performance assessment of concrete

127

Civil and Environmental Research, Vol.5 2013


Special Issue for International Congress on Materials & Structural Stability, Rabat, Morocco, 27-30 November 2013

[5] B.R. Ellingwood: Validation studies of seismic PRAs.


Nuclear Engineering and Design, 123 (1990),
p. 189196.

gravity dams structures are identified and also evaluated.


The sensitivity study presented utilizes the uncertain
parameters as inputs variables to identify which modeling
parameters significantly impact the seismic response
(output variable) of a number of different component
responses in concrete gravity dams. The main
conclusions from the presented comparison of seismic
vulnerability curves of concrete gravity dams under nearfault ground motions could be summarized as follows.

[6] B. G. Nielson, R. DesRoches: Seismic fragility


methodology for highway bridges using a
component level approach. Earthquake Engineering
and Structural Dynamics, 36 (2007), p. 823-839
[7] J. Zhang, Y. Huo, S. J. Brandenberg and P.
Kashighandi: Effects of structural characterizations
on fragility functions of bridges subject to seismic
shaking and lateral spreading. Earthquake
Engineering and Engineering Vibration Vol. 7
(2008), p. 369-382

It was found that, for all limits states LS1; LS2, LS3 and
LS4, the probability of failures is important for low
structural failure modes. However, this probability
decreases with the increase of these structural failure
modes. For limit state LS1, the probability of failure for
tensile stress is about 2 MPa. Limit state LS2 presented a
lower fragility for a sliding at dam-foundation interface
of 20 mm this value is 43.3 %. For the fragility of the
displacement at the top of the dam which is characterized
by the limit state LS3, it was found that the likelihood of
displacement is very important is about 82.45%. For limit
state LS4, it was found that there is no great risk; the
compressive stress is very small compared to the
compressive stress of concrete, therefore, there is no
significant risk at the heel of the dam. The results of
sensitivity analysis, however, have been found that the
variables parameters the cohesion (Var2); the young
modulus of concrete (Var4); the young modulus of soil
(Var5) and the compressive strength of concrete (Var6)
for both faults earthquakes have a considerable sensitivity
with high correlation on structure responses for
displacement at the top of dam and stress at the heel of
dam. However, a further study using various types of
dams must be necessary to draw a solid conclusion for
the fragility curves of dams structures.

[8] M. Sasan, A. Der Kiureghian and V. V. Bertero:


Seismic fragility of short period reinforced concrete
structural walls under near-source ground motions.
Structural Safety, 24 (2002), p. 123138
[9] M. A. Erberik, A. S. Elnashai: Fragility analysis of
flat-slab
structures.
Engineering
Structures,
26 (2004), p. 937948
[10] B. R. Ellingwood, O. C. Celik and K. Kinali:
Fragility assessment of building structural systems
in Mid-America. Earthquake Engineering and
Structural Dynamics, 36 (2007), p. 1935-1952
[11] E. M. Gneyisi, G. Altay: Seismic fragility
assessment of effectiveness of viscous dampers in
R/C buildings under scenario earthquakes.
Structural Safety, 30 (2008), p. 461-480
[12] D. M. Seyedi, P. Gehl, J. Douglas, L. Davenne, N.
Mezher and S. Ghavamian: Development of seismic
fragility surfaces for reinforced concrete buildings
by means of nonlinear time-history analysis.
Earthquake Engineering and Structural Dynamics,
39 (2010), p. 91-108
[13] O. C. Celik, B. R. Ellingwood: Seismic fragilities for
non-ductile reinforced concrete frames Role of
aleatoric and epistemic uncertainties. Structural
Safety, 32 (2010), p. 112

References
[1] M. HAMDACHE: Seismic Hazard Estimation in
Northern Algeria. Natural Hazards, 18 (1998),
p. 119144

[14] C. Kafali, M. Grigoriu: Seismic fragility analysis:


Application to simple linear and nonlinear
systems. Earthquake Engineering and Structural
Dynamics, 36 (2007), p. 18851900

[2] J. A. Pelaez Montilla, M. Hamdache and C. L.


Casado: Seismic hazard in Northern Algeria using
spatially smoothed seismicity. Results for peak
ground acceleration. Tectonophysics, 372 (2003),
p. 105 119

[15] B. Alavi, H. Krawinkler: Effects of Near-Fault


Ground Motions on Frame Structures. Department
of Civil and Environmental Engineering Stanford
University Report N 138, (2001)
[16] P. G. Somerville: Magnitude scaling of the near
fault rupture directivity pulse. Physics of the Earth
and Planetary Interiors, 137 (2003), p. 201212

[3] J. A. PELAEZ et al: Seismic Hazard in Terms of


Spectral Accelerations and Uniform Hazard Spectra
in Northern Algeria. Pure and Applied Geophysics,
163 (2006), p. 119135

[17] PEER, http://peer.berkeley.edu/svbin/GeneralSearch

[4] R.P. Kennedy, M.K. Ravindra: Seismic fragilities for


nuclear power plant studies. Nuclear Engineering
and Design, 79 (1984), p. 47 68

[18] B. R. Ellingwood: Earthquake risk for building


structures. Reliability Engineering and System
Safety, 74 (2001), p. 251-262

128

Civil and Environmental Research, Vol.5 2013


Special Issue for International Congress on Materials & Structural Stability, Rabat, Morocco, 27-30 November 2013

[19] J. G. MacGregor, S. A. Mirza, B. R. Ellingwood:


Statistical analysis of resistance of reinforced and
prestressed concrete members. ACI Journal, 80
(1983), p. 167176
[20] A. A. Afrouz: Practical Handbook of Rock Mass
Classification Systems and Modes of Ground
Failure. CRC Press: Boca Raton, FL, (1992)
[21] E. Hoek, P. K. Kaiser, W. F. Bawden: Support of
Underground Excavations in Hard Rock. A. A.
Balkema: Rotterdam, (1997)
[22]

Z. T. Bienawski: Engineering Rock


Classifications. Wiley: New York, (1989)

Mass

[23] B. Singh, R.K. Goel: Rock Mass Classification a


Practical Approach in Civil Engineering. Elsevier,
Oxford, (1999)
[24] M.D. McKay, W. J. Conover, R. J. Beckman: A
comparison of three methods for selecting values of
input variables in the analysis of output from a
computer code. Technometrics, 21 (1979), p23945
[25] A. Olsson, G. Sandberg, O. Dahlblom: On Latin
hypercube sampling for structural reliability
analysis. Structural Safety, 25 (2003), p. 4768
[26] J.C. Helton, F.J. Davis: Latin hypercube sampling
and the propagation of uncertainty in analyses of
complex systems. Reliability Engineering and
System Safety, 81 (2003), p. 2369.
[27] F. Lopez-Caballero, A. Modaressi-FarahmandRazavi:
Assessment of variability and
uncertainties effects on the seismic response of a
liquefiable soil profile. Soil Dynamics and
Earthquake Engineering, 30, Issue 7, (2010),
p 600613
[28] M. D. Bensalah, M. Bensaibi and A. Modaressi:
Assessment of the Torsion Effect in Asymmetric
Buildings under Seismic Loading. Applied
Mechanics and Materials, 256-259 (2013),
p. 2222-2228

129